kern_subr.c revision 1.165 1 /* $NetBSD: kern_subr.c,v 1.165 2007/10/12 13:00:18 ad Exp $ */
2
3 /*-
4 * Copyright (c) 1997, 1998, 1999, 2002, 2007, 2006 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
9 * NASA Ames Research Center, and by Luke Mewburn.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. All advertising materials mentioning features or use of this software
20 * must display the following acknowledgement:
21 * This product includes software developed by the NetBSD
22 * Foundation, Inc. and its contributors.
23 * 4. Neither the name of The NetBSD Foundation nor the names of its
24 * contributors may be used to endorse or promote products derived
25 * from this software without specific prior written permission.
26 *
27 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37 * POSSIBILITY OF SUCH DAMAGE.
38 */
39
40 /*
41 * Copyright (c) 1982, 1986, 1991, 1993
42 * The Regents of the University of California. All rights reserved.
43 * (c) UNIX System Laboratories, Inc.
44 * All or some portions of this file are derived from material licensed
45 * to the University of California by American Telephone and Telegraph
46 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
47 * the permission of UNIX System Laboratories, Inc.
48 *
49 * Copyright (c) 1992, 1993
50 * The Regents of the University of California. All rights reserved.
51 *
52 * This software was developed by the Computer Systems Engineering group
53 * at Lawrence Berkeley Laboratory under DARPA contract BG 91-66 and
54 * contributed to Berkeley.
55 *
56 * All advertising materials mentioning features or use of this software
57 * must display the following acknowledgement:
58 * This product includes software developed by the University of
59 * California, Lawrence Berkeley Laboratory.
60 *
61 * Redistribution and use in source and binary forms, with or without
62 * modification, are permitted provided that the following conditions
63 * are met:
64 * 1. Redistributions of source code must retain the above copyright
65 * notice, this list of conditions and the following disclaimer.
66 * 2. Redistributions in binary form must reproduce the above copyright
67 * notice, this list of conditions and the following disclaimer in the
68 * documentation and/or other materials provided with the distribution.
69 * 3. Neither the name of the University nor the names of its contributors
70 * may be used to endorse or promote products derived from this software
71 * without specific prior written permission.
72 *
73 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
74 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
75 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
76 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
77 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
78 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
79 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
80 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
81 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
82 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
83 * SUCH DAMAGE.
84 *
85 * @(#)kern_subr.c 8.4 (Berkeley) 2/14/95
86 */
87
88 #include <sys/cdefs.h>
89 __KERNEL_RCSID(0, "$NetBSD: kern_subr.c,v 1.165 2007/10/12 13:00:18 ad Exp $");
90
91 #include "opt_ddb.h"
92 #include "opt_md.h"
93 #include "opt_syscall_debug.h"
94 #include "opt_ktrace.h"
95 #include "opt_ptrace.h"
96 #include "opt_systrace.h"
97 #include "opt_powerhook.h"
98 #include "opt_tftproot.h"
99
100 #include <sys/param.h>
101 #include <sys/systm.h>
102 #include <sys/proc.h>
103 #include <sys/malloc.h>
104 #include <sys/mount.h>
105 #include <sys/device.h>
106 #include <sys/reboot.h>
107 #include <sys/conf.h>
108 #include <sys/disk.h>
109 #include <sys/disklabel.h>
110 #include <sys/queue.h>
111 #include <sys/systrace.h>
112 #include <sys/ktrace.h>
113 #include <sys/ptrace.h>
114 #include <sys/fcntl.h>
115 #include <sys/kauth.h>
116 #include <sys/vnode.h>
117
118 #include <uvm/uvm_extern.h>
119
120 #include <dev/cons.h>
121
122 #include <net/if.h>
123
124 /* XXX these should eventually move to subr_autoconf.c */
125 static struct device *finddevice(const char *);
126 static struct device *getdisk(char *, int, int, dev_t *, int);
127 static struct device *parsedisk(char *, int, int, dev_t *);
128 static const char *getwedgename(const char *, int);
129
130 /*
131 * A generic linear hook.
132 */
133 struct hook_desc {
134 LIST_ENTRY(hook_desc) hk_list;
135 void (*hk_fn)(void *);
136 void *hk_arg;
137 };
138 typedef LIST_HEAD(, hook_desc) hook_list_t;
139
140 MALLOC_DEFINE(M_IOV, "iov", "large iov's");
141
142 #ifdef TFTPROOT
143 int tftproot_dhcpboot(struct device *);
144 #endif
145
146 void
147 uio_setup_sysspace(struct uio *uio)
148 {
149
150 uio->uio_vmspace = vmspace_kernel();
151 }
152
153 int
154 uiomove(void *buf, size_t n, struct uio *uio)
155 {
156 struct vmspace *vm = uio->uio_vmspace;
157 struct iovec *iov;
158 u_int cnt;
159 int error = 0;
160 size_t on;
161 char *cp = buf;
162 #ifdef MULTIPROCESSOR
163 int hold_count;
164 #endif
165
166 if ((on = n) >= 1024) {
167 KERNEL_UNLOCK_ALL(NULL, &hold_count);
168 }
169
170 ASSERT_SLEEPABLE(NULL, "uiomove");
171
172 #ifdef DIAGNOSTIC
173 if (uio->uio_rw != UIO_READ && uio->uio_rw != UIO_WRITE)
174 panic("uiomove: mode");
175 #endif
176 while (n > 0 && uio->uio_resid) {
177 iov = uio->uio_iov;
178 cnt = iov->iov_len;
179 if (cnt == 0) {
180 KASSERT(uio->uio_iovcnt > 0);
181 uio->uio_iov++;
182 uio->uio_iovcnt--;
183 continue;
184 }
185 if (cnt > n)
186 cnt = n;
187 if (!VMSPACE_IS_KERNEL_P(vm)) {
188 if (curcpu()->ci_schedstate.spc_flags &
189 SPCF_SHOULDYIELD)
190 preempt();
191 }
192
193 if (uio->uio_rw == UIO_READ) {
194 error = copyout_vmspace(vm, cp, iov->iov_base,
195 cnt);
196 } else {
197 error = copyin_vmspace(vm, iov->iov_base, cp,
198 cnt);
199 }
200 if (error) {
201 break;
202 }
203 iov->iov_base = (char *)iov->iov_base + cnt;
204 iov->iov_len -= cnt;
205 uio->uio_resid -= cnt;
206 uio->uio_offset += cnt;
207 cp += cnt;
208 KDASSERT(cnt <= n);
209 n -= cnt;
210 }
211
212 if (on >= 1024) {
213 KERNEL_LOCK(hold_count, NULL);
214 }
215 return (error);
216 }
217
218 /*
219 * Wrapper for uiomove() that validates the arguments against a known-good
220 * kernel buffer.
221 */
222 int
223 uiomove_frombuf(void *buf, size_t buflen, struct uio *uio)
224 {
225 size_t offset;
226
227 if (uio->uio_offset < 0 || /* uio->uio_resid < 0 || */
228 (offset = uio->uio_offset) != uio->uio_offset)
229 return (EINVAL);
230 if (offset >= buflen)
231 return (0);
232 return (uiomove((char *)buf + offset, buflen - offset, uio));
233 }
234
235 /*
236 * Give next character to user as result of read.
237 */
238 int
239 ureadc(int c, struct uio *uio)
240 {
241 struct iovec *iov;
242
243 if (uio->uio_resid <= 0)
244 panic("ureadc: non-positive resid");
245 again:
246 if (uio->uio_iovcnt <= 0)
247 panic("ureadc: non-positive iovcnt");
248 iov = uio->uio_iov;
249 if (iov->iov_len <= 0) {
250 uio->uio_iovcnt--;
251 uio->uio_iov++;
252 goto again;
253 }
254 if (!VMSPACE_IS_KERNEL_P(uio->uio_vmspace)) {
255 if (subyte(iov->iov_base, c) < 0)
256 return (EFAULT);
257 } else {
258 *(char *)iov->iov_base = c;
259 }
260 iov->iov_base = (char *)iov->iov_base + 1;
261 iov->iov_len--;
262 uio->uio_resid--;
263 uio->uio_offset++;
264 return (0);
265 }
266
267 /*
268 * Like copyin(), but operates on an arbitrary vmspace.
269 */
270 int
271 copyin_vmspace(struct vmspace *vm, const void *uaddr, void *kaddr, size_t len)
272 {
273 struct iovec iov;
274 struct uio uio;
275 int error;
276
277 if (len == 0)
278 return (0);
279
280 if (VMSPACE_IS_KERNEL_P(vm)) {
281 return kcopy(uaddr, kaddr, len);
282 }
283 if (__predict_true(vm == curproc->p_vmspace)) {
284 return copyin(uaddr, kaddr, len);
285 }
286
287 iov.iov_base = kaddr;
288 iov.iov_len = len;
289 uio.uio_iov = &iov;
290 uio.uio_iovcnt = 1;
291 uio.uio_offset = (off_t)(intptr_t)uaddr;
292 uio.uio_resid = len;
293 uio.uio_rw = UIO_READ;
294 UIO_SETUP_SYSSPACE(&uio);
295 error = uvm_io(&vm->vm_map, &uio);
296
297 return (error);
298 }
299
300 /*
301 * Like copyout(), but operates on an arbitrary vmspace.
302 */
303 int
304 copyout_vmspace(struct vmspace *vm, const void *kaddr, void *uaddr, size_t len)
305 {
306 struct iovec iov;
307 struct uio uio;
308 int error;
309
310 if (len == 0)
311 return (0);
312
313 if (VMSPACE_IS_KERNEL_P(vm)) {
314 return kcopy(kaddr, uaddr, len);
315 }
316 if (__predict_true(vm == curproc->p_vmspace)) {
317 return copyout(kaddr, uaddr, len);
318 }
319
320 iov.iov_base = __UNCONST(kaddr); /* XXXUNCONST cast away const */
321 iov.iov_len = len;
322 uio.uio_iov = &iov;
323 uio.uio_iovcnt = 1;
324 uio.uio_offset = (off_t)(intptr_t)uaddr;
325 uio.uio_resid = len;
326 uio.uio_rw = UIO_WRITE;
327 UIO_SETUP_SYSSPACE(&uio);
328 error = uvm_io(&vm->vm_map, &uio);
329
330 return (error);
331 }
332
333 /*
334 * Like copyin(), but operates on an arbitrary process.
335 */
336 int
337 copyin_proc(struct proc *p, const void *uaddr, void *kaddr, size_t len)
338 {
339 struct vmspace *vm;
340 int error;
341
342 error = proc_vmspace_getref(p, &vm);
343 if (error) {
344 return error;
345 }
346 error = copyin_vmspace(vm, uaddr, kaddr, len);
347 uvmspace_free(vm);
348
349 return error;
350 }
351
352 /*
353 * Like copyout(), but operates on an arbitrary process.
354 */
355 int
356 copyout_proc(struct proc *p, const void *kaddr, void *uaddr, size_t len)
357 {
358 struct vmspace *vm;
359 int error;
360
361 error = proc_vmspace_getref(p, &vm);
362 if (error) {
363 return error;
364 }
365 error = copyout_vmspace(vm, kaddr, uaddr, len);
366 uvmspace_free(vm);
367
368 return error;
369 }
370
371 /*
372 * Like copyin(), except it operates on kernel addresses when the FKIOCTL
373 * flag is passed in `ioctlflags' from the ioctl call.
374 */
375 int
376 ioctl_copyin(int ioctlflags, const void *src, void *dst, size_t len)
377 {
378 if (ioctlflags & FKIOCTL)
379 return kcopy(src, dst, len);
380 return copyin(src, dst, len);
381 }
382
383 /*
384 * Like copyout(), except it operates on kernel addresses when the FKIOCTL
385 * flag is passed in `ioctlflags' from the ioctl call.
386 */
387 int
388 ioctl_copyout(int ioctlflags, const void *src, void *dst, size_t len)
389 {
390 if (ioctlflags & FKIOCTL)
391 return kcopy(src, dst, len);
392 return copyout(src, dst, len);
393 }
394
395 static void *
396 hook_establish(hook_list_t *list, void (*fn)(void *), void *arg)
397 {
398 struct hook_desc *hd;
399
400 hd = malloc(sizeof(*hd), M_DEVBUF, M_NOWAIT);
401 if (hd == NULL)
402 return (NULL);
403
404 hd->hk_fn = fn;
405 hd->hk_arg = arg;
406 LIST_INSERT_HEAD(list, hd, hk_list);
407
408 return (hd);
409 }
410
411 static void
412 hook_disestablish(hook_list_t *list, void *vhook)
413 {
414 #ifdef DIAGNOSTIC
415 struct hook_desc *hd;
416
417 LIST_FOREACH(hd, list, hk_list) {
418 if (hd == vhook)
419 break;
420 }
421
422 if (hd == NULL)
423 panic("hook_disestablish: hook %p not established", vhook);
424 #endif
425 LIST_REMOVE((struct hook_desc *)vhook, hk_list);
426 free(vhook, M_DEVBUF);
427 }
428
429 static void
430 hook_destroy(hook_list_t *list)
431 {
432 struct hook_desc *hd;
433
434 while ((hd = LIST_FIRST(list)) != NULL) {
435 LIST_REMOVE(hd, hk_list);
436 free(hd, M_DEVBUF);
437 }
438 }
439
440 static void
441 hook_proc_run(hook_list_t *list, struct proc *p)
442 {
443 struct hook_desc *hd;
444
445 for (hd = LIST_FIRST(list); hd != NULL; hd = LIST_NEXT(hd, hk_list)) {
446 ((void (*)(struct proc *, void *))*hd->hk_fn)(p,
447 hd->hk_arg);
448 }
449 }
450
451 /*
452 * "Shutdown hook" types, functions, and variables.
453 *
454 * Should be invoked immediately before the
455 * system is halted or rebooted, i.e. after file systems unmounted,
456 * after crash dump done, etc.
457 *
458 * Each shutdown hook is removed from the list before it's run, so that
459 * it won't be run again.
460 */
461
462 static hook_list_t shutdownhook_list;
463
464 void *
465 shutdownhook_establish(void (*fn)(void *), void *arg)
466 {
467 return hook_establish(&shutdownhook_list, fn, arg);
468 }
469
470 void
471 shutdownhook_disestablish(void *vhook)
472 {
473 hook_disestablish(&shutdownhook_list, vhook);
474 }
475
476 /*
477 * Run shutdown hooks. Should be invoked immediately before the
478 * system is halted or rebooted, i.e. after file systems unmounted,
479 * after crash dump done, etc.
480 *
481 * Each shutdown hook is removed from the list before it's run, so that
482 * it won't be run again.
483 */
484 void
485 doshutdownhooks(void)
486 {
487 struct hook_desc *dp;
488
489 while ((dp = LIST_FIRST(&shutdownhook_list)) != NULL) {
490 LIST_REMOVE(dp, hk_list);
491 (*dp->hk_fn)(dp->hk_arg);
492 #if 0
493 /*
494 * Don't bother freeing the hook structure,, since we may
495 * be rebooting because of a memory corruption problem,
496 * and this might only make things worse. It doesn't
497 * matter, anyway, since the system is just about to
498 * reboot.
499 */
500 free(dp, M_DEVBUF);
501 #endif
502 }
503 }
504
505 /*
506 * "Mountroot hook" types, functions, and variables.
507 */
508
509 static hook_list_t mountroothook_list;
510
511 void *
512 mountroothook_establish(void (*fn)(struct device *), struct device *dev)
513 {
514 return hook_establish(&mountroothook_list, (void (*)(void *))fn, dev);
515 }
516
517 void
518 mountroothook_disestablish(void *vhook)
519 {
520 hook_disestablish(&mountroothook_list, vhook);
521 }
522
523 void
524 mountroothook_destroy(void)
525 {
526 hook_destroy(&mountroothook_list);
527 }
528
529 void
530 domountroothook(void)
531 {
532 struct hook_desc *hd;
533
534 LIST_FOREACH(hd, &mountroothook_list, hk_list) {
535 if (hd->hk_arg == (void *)root_device) {
536 (*hd->hk_fn)(hd->hk_arg);
537 return;
538 }
539 }
540 }
541
542 static hook_list_t exechook_list;
543
544 void *
545 exechook_establish(void (*fn)(struct proc *, void *), void *arg)
546 {
547 return hook_establish(&exechook_list, (void (*)(void *))fn, arg);
548 }
549
550 void
551 exechook_disestablish(void *vhook)
552 {
553 hook_disestablish(&exechook_list, vhook);
554 }
555
556 /*
557 * Run exec hooks.
558 */
559 void
560 doexechooks(struct proc *p)
561 {
562 hook_proc_run(&exechook_list, p);
563 }
564
565 static hook_list_t exithook_list;
566
567 void *
568 exithook_establish(void (*fn)(struct proc *, void *), void *arg)
569 {
570 return hook_establish(&exithook_list, (void (*)(void *))fn, arg);
571 }
572
573 void
574 exithook_disestablish(void *vhook)
575 {
576 hook_disestablish(&exithook_list, vhook);
577 }
578
579 /*
580 * Run exit hooks.
581 */
582 void
583 doexithooks(struct proc *p)
584 {
585 hook_proc_run(&exithook_list, p);
586 }
587
588 static hook_list_t forkhook_list;
589
590 void *
591 forkhook_establish(void (*fn)(struct proc *, struct proc *))
592 {
593 return hook_establish(&forkhook_list, (void (*)(void *))fn, NULL);
594 }
595
596 void
597 forkhook_disestablish(void *vhook)
598 {
599 hook_disestablish(&forkhook_list, vhook);
600 }
601
602 /*
603 * Run fork hooks.
604 */
605 void
606 doforkhooks(struct proc *p2, struct proc *p1)
607 {
608 struct hook_desc *hd;
609
610 LIST_FOREACH(hd, &forkhook_list, hk_list) {
611 ((void (*)(struct proc *, struct proc *))*hd->hk_fn)
612 (p2, p1);
613 }
614 }
615
616 /*
617 * "Power hook" types, functions, and variables.
618 * The list of power hooks is kept ordered with the last registered hook
619 * first.
620 * When running the hooks on power down the hooks are called in reverse
621 * registration order, when powering up in registration order.
622 */
623 struct powerhook_desc {
624 CIRCLEQ_ENTRY(powerhook_desc) sfd_list;
625 void (*sfd_fn)(int, void *);
626 void *sfd_arg;
627 char sfd_name[16];
628 };
629
630 static CIRCLEQ_HEAD(, powerhook_desc) powerhook_list =
631 CIRCLEQ_HEAD_INITIALIZER(powerhook_list);
632
633 void *
634 powerhook_establish(const char *name, void (*fn)(int, void *), void *arg)
635 {
636 struct powerhook_desc *ndp;
637
638 ndp = (struct powerhook_desc *)
639 malloc(sizeof(*ndp), M_DEVBUF, M_NOWAIT);
640 if (ndp == NULL)
641 return (NULL);
642
643 ndp->sfd_fn = fn;
644 ndp->sfd_arg = arg;
645 strlcpy(ndp->sfd_name, name, sizeof(ndp->sfd_name));
646 CIRCLEQ_INSERT_HEAD(&powerhook_list, ndp, sfd_list);
647
648 return (ndp);
649 }
650
651 void
652 powerhook_disestablish(void *vhook)
653 {
654 #ifdef DIAGNOSTIC
655 struct powerhook_desc *dp;
656
657 CIRCLEQ_FOREACH(dp, &powerhook_list, sfd_list)
658 if (dp == vhook)
659 goto found;
660 panic("powerhook_disestablish: hook %p not established", vhook);
661 found:
662 #endif
663
664 CIRCLEQ_REMOVE(&powerhook_list, (struct powerhook_desc *)vhook,
665 sfd_list);
666 free(vhook, M_DEVBUF);
667 }
668
669 /*
670 * Run power hooks.
671 */
672 void
673 dopowerhooks(int why)
674 {
675 struct powerhook_desc *dp;
676
677 #ifdef POWERHOOK_DEBUG
678 const char *why_name;
679 static const char * pwr_names[] = {PWR_NAMES};
680 why_name = why < __arraycount(pwr_names) ? pwr_names[why] : "???";
681 #endif
682
683 if (why == PWR_RESUME || why == PWR_SOFTRESUME) {
684 CIRCLEQ_FOREACH_REVERSE(dp, &powerhook_list, sfd_list) {
685 #ifdef POWERHOOK_DEBUG
686 printf("dopowerhooks %s: %s (%p)\n", why_name, dp->sfd_name, dp);
687 #endif
688 (*dp->sfd_fn)(why, dp->sfd_arg);
689 }
690 } else {
691 CIRCLEQ_FOREACH(dp, &powerhook_list, sfd_list) {
692 #ifdef POWERHOOK_DEBUG
693 printf("dopowerhooks %s: %s (%p)\n", why_name, dp->sfd_name, dp);
694 #endif
695 (*dp->sfd_fn)(why, dp->sfd_arg);
696 }
697 }
698
699 #ifdef POWERHOOK_DEBUG
700 printf("dopowerhooks: %s done\n", why_name);
701 #endif
702 }
703
704 static int
705 isswap(struct device *dv)
706 {
707 struct dkwedge_info wi;
708 struct vnode *vn;
709 int error;
710
711 if (device_class(dv) != DV_DISK || !device_is_a(dv, "dk"))
712 return 0;
713
714 if ((vn = opendisk(dv)) == NULL)
715 return 0;
716
717 error = VOP_IOCTL(vn, DIOCGWEDGEINFO, &wi, FREAD, NOCRED, 0);
718 VOP_CLOSE(vn, FREAD, NOCRED, 0);
719 vput(vn);
720 if (error) {
721 #ifdef DEBUG_WEDGE
722 printf("%s: Get wedge info returned %d\n", dv->dv_xname, error);
723 #endif
724 return 0;
725 }
726 return strcmp(wi.dkw_ptype, DKW_PTYPE_SWAP) == 0;
727 }
728
729 /*
730 * Determine the root device and, if instructed to, the root file system.
731 */
732
733 #include "md.h"
734 #if NMD == 0
735 #undef MEMORY_DISK_HOOKS
736 #endif
737
738 #ifdef MEMORY_DISK_HOOKS
739 static struct device fakemdrootdev[NMD];
740 extern struct cfdriver md_cd;
741 #endif
742
743 #ifdef MEMORY_DISK_IS_ROOT
744 #define BOOT_FROM_MEMORY_HOOKS 1
745 #endif
746
747 /*
748 * The device and wedge that we booted from. If booted_wedge is NULL,
749 * the we might consult booted_partition.
750 */
751 struct device *booted_device;
752 struct device *booted_wedge;
753 int booted_partition;
754
755 /*
756 * Use partition letters if it's a disk class but not a wedge.
757 * XXX Check for wedge is kinda gross.
758 */
759 #define DEV_USES_PARTITIONS(dv) \
760 (device_class((dv)) == DV_DISK && \
761 !device_is_a((dv), "dk"))
762
763 void
764 setroot(struct device *bootdv, int bootpartition)
765 {
766 struct device *dv;
767 int len, majdev;
768 #ifdef MEMORY_DISK_HOOKS
769 int i;
770 #endif
771 dev_t nrootdev;
772 dev_t ndumpdev = NODEV;
773 char buf[128];
774 const char *rootdevname;
775 const char *dumpdevname;
776 struct device *rootdv = NULL; /* XXX gcc -Wuninitialized */
777 struct device *dumpdv = NULL;
778 struct ifnet *ifp;
779 const char *deffsname;
780 struct vfsops *vops;
781
782 #ifdef TFTPROOT
783 if (tftproot_dhcpboot(bootdv) != 0)
784 boothowto |= RB_ASKNAME;
785 #endif
786
787 #ifdef MEMORY_DISK_HOOKS
788 for (i = 0; i < NMD; i++) {
789 fakemdrootdev[i].dv_class = DV_DISK;
790 fakemdrootdev[i].dv_cfdata = NULL;
791 fakemdrootdev[i].dv_cfdriver = &md_cd;
792 fakemdrootdev[i].dv_unit = i;
793 fakemdrootdev[i].dv_parent = NULL;
794 snprintf(fakemdrootdev[i].dv_xname,
795 sizeof(fakemdrootdev[i].dv_xname), "md%d", i);
796 }
797 #endif /* MEMORY_DISK_HOOKS */
798
799 #ifdef MEMORY_DISK_IS_ROOT
800 bootdv = &fakemdrootdev[0];
801 bootpartition = 0;
802 #endif
803
804 /*
805 * If NFS is specified as the file system, and we found
806 * a DV_DISK boot device (or no boot device at all), then
807 * find a reasonable network interface for "rootspec".
808 */
809 vops = vfs_getopsbyname("nfs");
810 if (vops != NULL && vops->vfs_mountroot == mountroot &&
811 rootspec == NULL &&
812 (bootdv == NULL || device_class(bootdv) != DV_IFNET)) {
813 IFNET_FOREACH(ifp) {
814 if ((ifp->if_flags &
815 (IFF_LOOPBACK|IFF_POINTOPOINT)) == 0)
816 break;
817 }
818 if (ifp == NULL) {
819 /*
820 * Can't find a suitable interface; ask the
821 * user.
822 */
823 boothowto |= RB_ASKNAME;
824 } else {
825 /*
826 * Have a suitable interface; behave as if
827 * the user specified this interface.
828 */
829 rootspec = (const char *)ifp->if_xname;
830 }
831 }
832 if (vops != NULL)
833 vfs_delref(vops);
834
835 /*
836 * If wildcarded root and we the boot device wasn't determined,
837 * ask the user.
838 */
839 if (rootspec == NULL && bootdv == NULL)
840 boothowto |= RB_ASKNAME;
841
842 top:
843 if (boothowto & RB_ASKNAME) {
844 struct device *defdumpdv;
845
846 for (;;) {
847 printf("root device");
848 if (bootdv != NULL) {
849 printf(" (default %s", bootdv->dv_xname);
850 if (DEV_USES_PARTITIONS(bootdv))
851 printf("%c", bootpartition + 'a');
852 printf(")");
853 }
854 printf(": ");
855 len = cngetsn(buf, sizeof(buf));
856 if (len == 0 && bootdv != NULL) {
857 strlcpy(buf, bootdv->dv_xname, sizeof(buf));
858 len = strlen(buf);
859 }
860 if (len > 0 && buf[len - 1] == '*') {
861 buf[--len] = '\0';
862 dv = getdisk(buf, len, 1, &nrootdev, 0);
863 if (dv != NULL) {
864 rootdv = dv;
865 break;
866 }
867 }
868 dv = getdisk(buf, len, bootpartition, &nrootdev, 0);
869 if (dv != NULL) {
870 rootdv = dv;
871 break;
872 }
873 }
874
875 /*
876 * Set up the default dump device. If root is on
877 * a network device, there is no default dump
878 * device, since we don't support dumps to the
879 * network.
880 */
881 if (DEV_USES_PARTITIONS(rootdv) == 0)
882 defdumpdv = NULL;
883 else
884 defdumpdv = rootdv;
885
886 for (;;) {
887 printf("dump device");
888 if (defdumpdv != NULL) {
889 /*
890 * Note, we know it's a disk if we get here.
891 */
892 printf(" (default %sb)", defdumpdv->dv_xname);
893 }
894 printf(": ");
895 len = cngetsn(buf, sizeof(buf));
896 if (len == 0) {
897 if (defdumpdv != NULL) {
898 ndumpdev = MAKEDISKDEV(major(nrootdev),
899 DISKUNIT(nrootdev), 1);
900 }
901 dumpdv = defdumpdv;
902 break;
903 }
904 if (len == 4 && strcmp(buf, "none") == 0) {
905 dumpdv = NULL;
906 break;
907 }
908 dv = getdisk(buf, len, 1, &ndumpdev, 1);
909 if (dv != NULL) {
910 dumpdv = dv;
911 break;
912 }
913 }
914
915 rootdev = nrootdev;
916 dumpdev = ndumpdev;
917
918 for (vops = LIST_FIRST(&vfs_list); vops != NULL;
919 vops = LIST_NEXT(vops, vfs_list)) {
920 if (vops->vfs_mountroot != NULL &&
921 vops->vfs_mountroot == mountroot)
922 break;
923 }
924
925 if (vops == NULL) {
926 mountroot = NULL;
927 deffsname = "generic";
928 } else
929 deffsname = vops->vfs_name;
930
931 for (;;) {
932 printf("file system (default %s): ", deffsname);
933 len = cngetsn(buf, sizeof(buf));
934 if (len == 0)
935 break;
936 if (len == 4 && strcmp(buf, "halt") == 0)
937 cpu_reboot(RB_HALT, NULL);
938 else if (len == 6 && strcmp(buf, "reboot") == 0)
939 cpu_reboot(0, NULL);
940 #if defined(DDB)
941 else if (len == 3 && strcmp(buf, "ddb") == 0) {
942 console_debugger();
943 }
944 #endif
945 else if (len == 7 && strcmp(buf, "generic") == 0) {
946 mountroot = NULL;
947 break;
948 }
949 vops = vfs_getopsbyname(buf);
950 if (vops == NULL || vops->vfs_mountroot == NULL) {
951 printf("use one of: generic");
952 for (vops = LIST_FIRST(&vfs_list);
953 vops != NULL;
954 vops = LIST_NEXT(vops, vfs_list)) {
955 if (vops->vfs_mountroot != NULL)
956 printf(" %s", vops->vfs_name);
957 }
958 #if defined(DDB)
959 printf(" ddb");
960 #endif
961 printf(" halt reboot\n");
962 } else {
963 mountroot = vops->vfs_mountroot;
964 vfs_delref(vops);
965 break;
966 }
967 }
968
969 } else if (rootspec == NULL) {
970 /*
971 * Wildcarded root; use the boot device.
972 */
973 rootdv = bootdv;
974
975 majdev = devsw_name2blk(bootdv->dv_xname, NULL, 0);
976 if (majdev >= 0) {
977 /*
978 * Root is on a disk. `bootpartition' is root,
979 * unless the device does not use partitions.
980 */
981 if (DEV_USES_PARTITIONS(bootdv))
982 rootdev = MAKEDISKDEV(majdev,
983 device_unit(bootdv),
984 bootpartition);
985 else
986 rootdev = makedev(majdev, device_unit(bootdv));
987 }
988 } else {
989
990 /*
991 * `root on <dev> ...'
992 */
993
994 /*
995 * If it's a network interface, we can bail out
996 * early.
997 */
998 dv = finddevice(rootspec);
999 if (dv != NULL && device_class(dv) == DV_IFNET) {
1000 rootdv = dv;
1001 goto haveroot;
1002 }
1003
1004 if (rootdev == NODEV &&
1005 device_class(dv) == DV_DISK && device_is_a(dv, "dk") &&
1006 (majdev = devsw_name2blk(dv->dv_xname, NULL, 0)) >= 0)
1007 rootdev = makedev(majdev, device_unit(dv));
1008
1009 rootdevname = devsw_blk2name(major(rootdev));
1010 if (rootdevname == NULL) {
1011 printf("unknown device major 0x%x\n", rootdev);
1012 boothowto |= RB_ASKNAME;
1013 goto top;
1014 }
1015 memset(buf, 0, sizeof(buf));
1016 snprintf(buf, sizeof(buf), "%s%d", rootdevname,
1017 DISKUNIT(rootdev));
1018
1019 rootdv = finddevice(buf);
1020 if (rootdv == NULL) {
1021 printf("device %s (0x%x) not configured\n",
1022 buf, rootdev);
1023 boothowto |= RB_ASKNAME;
1024 goto top;
1025 }
1026 }
1027
1028 haveroot:
1029
1030 root_device = rootdv;
1031
1032 switch (device_class(rootdv)) {
1033 case DV_IFNET:
1034 case DV_DISK:
1035 aprint_normal("root on %s", rootdv->dv_xname);
1036 if (DEV_USES_PARTITIONS(rootdv))
1037 aprint_normal("%c", DISKPART(rootdev) + 'a');
1038 break;
1039
1040 default:
1041 printf("can't determine root device\n");
1042 boothowto |= RB_ASKNAME;
1043 goto top;
1044 }
1045
1046 /*
1047 * Now configure the dump device.
1048 *
1049 * If we haven't figured out the dump device, do so, with
1050 * the following rules:
1051 *
1052 * (a) We already know dumpdv in the RB_ASKNAME case.
1053 *
1054 * (b) If dumpspec is set, try to use it. If the device
1055 * is not available, punt.
1056 *
1057 * (c) If dumpspec is not set, the dump device is
1058 * wildcarded or unspecified. If the root device
1059 * is DV_IFNET, punt. Otherwise, use partition b
1060 * of the root device.
1061 */
1062
1063 if (boothowto & RB_ASKNAME) { /* (a) */
1064 if (dumpdv == NULL)
1065 goto nodumpdev;
1066 } else if (dumpspec != NULL) { /* (b) */
1067 if (strcmp(dumpspec, "none") == 0 || dumpdev == NODEV) {
1068 /*
1069 * Operator doesn't want a dump device.
1070 * Or looks like they tried to pick a network
1071 * device. Oops.
1072 */
1073 goto nodumpdev;
1074 }
1075
1076 dumpdevname = devsw_blk2name(major(dumpdev));
1077 if (dumpdevname == NULL)
1078 goto nodumpdev;
1079 memset(buf, 0, sizeof(buf));
1080 snprintf(buf, sizeof(buf), "%s%d", dumpdevname,
1081 DISKUNIT(dumpdev));
1082
1083 dumpdv = finddevice(buf);
1084 if (dumpdv == NULL) {
1085 /*
1086 * Device not configured.
1087 */
1088 goto nodumpdev;
1089 }
1090 } else { /* (c) */
1091 if (DEV_USES_PARTITIONS(rootdv) == 0) {
1092 for (dv = TAILQ_FIRST(&alldevs); dv != NULL;
1093 dv = TAILQ_NEXT(dv, dv_list))
1094 if (isswap(dv))
1095 break;
1096 if (dv == NULL)
1097 goto nodumpdev;
1098
1099 majdev = devsw_name2blk(dv->dv_xname, NULL, 0);
1100 if (majdev < 0)
1101 goto nodumpdev;
1102 dumpdv = dv;
1103 dumpdev = makedev(majdev, device_unit(dumpdv));
1104 } else {
1105 dumpdv = rootdv;
1106 dumpdev = MAKEDISKDEV(major(rootdev),
1107 device_unit(dumpdv), 1);
1108 }
1109 }
1110
1111 aprint_normal(" dumps on %s", dumpdv->dv_xname);
1112 if (DEV_USES_PARTITIONS(dumpdv))
1113 aprint_normal("%c", DISKPART(dumpdev) + 'a');
1114 aprint_normal("\n");
1115 return;
1116
1117 nodumpdev:
1118 dumpdev = NODEV;
1119 aprint_normal("\n");
1120 }
1121
1122 static struct device *
1123 finddevice(const char *name)
1124 {
1125 const char *wname;
1126 struct device *dv;
1127 #if defined(BOOT_FROM_MEMORY_HOOKS)
1128 int j;
1129 #endif /* BOOT_FROM_MEMORY_HOOKS */
1130
1131 if ((wname = getwedgename(name, strlen(name))) != NULL)
1132 return dkwedge_find_by_wname(wname);
1133
1134 #ifdef BOOT_FROM_MEMORY_HOOKS
1135 for (j = 0; j < NMD; j++) {
1136 if (strcmp(name, fakemdrootdev[j].dv_xname) == 0)
1137 return &fakemdrootdev[j];
1138 }
1139 #endif /* BOOT_FROM_MEMORY_HOOKS */
1140
1141 TAILQ_FOREACH(dv, &alldevs, dv_list) {
1142 if (strcmp(dv->dv_xname, name) == 0)
1143 break;
1144 }
1145 return dv;
1146 }
1147
1148 static struct device *
1149 getdisk(char *str, int len, int defpart, dev_t *devp, int isdump)
1150 {
1151 struct device *dv;
1152 #ifdef MEMORY_DISK_HOOKS
1153 int i;
1154 #endif
1155
1156 if ((dv = parsedisk(str, len, defpart, devp)) == NULL) {
1157 printf("use one of:");
1158 #ifdef MEMORY_DISK_HOOKS
1159 if (isdump == 0)
1160 for (i = 0; i < NMD; i++)
1161 printf(" %s[a-%c]", fakemdrootdev[i].dv_xname,
1162 'a' + MAXPARTITIONS - 1);
1163 #endif
1164 TAILQ_FOREACH(dv, &alldevs, dv_list) {
1165 if (DEV_USES_PARTITIONS(dv))
1166 printf(" %s[a-%c]", dv->dv_xname,
1167 'a' + MAXPARTITIONS - 1);
1168 else if (device_class(dv) == DV_DISK)
1169 printf(" %s", dv->dv_xname);
1170 if (isdump == 0 && device_class(dv) == DV_IFNET)
1171 printf(" %s", dv->dv_xname);
1172 }
1173 dkwedge_print_wnames();
1174 if (isdump)
1175 printf(" none");
1176 #if defined(DDB)
1177 printf(" ddb");
1178 #endif
1179 printf(" halt reboot\n");
1180 }
1181 return dv;
1182 }
1183
1184 static const char *
1185 getwedgename(const char *name, int namelen)
1186 {
1187 const char *wpfx = "wedge:";
1188 const int wpfxlen = strlen(wpfx);
1189
1190 if (namelen < wpfxlen || strncmp(name, wpfx, wpfxlen) != 0)
1191 return NULL;
1192
1193 return name + wpfxlen;
1194 }
1195
1196 static struct device *
1197 parsedisk(char *str, int len, int defpart, dev_t *devp)
1198 {
1199 struct device *dv;
1200 const char *wname;
1201 char *cp, c;
1202 int majdev, part;
1203 #ifdef MEMORY_DISK_HOOKS
1204 int i;
1205 #endif
1206 if (len == 0)
1207 return (NULL);
1208
1209 if (len == 4 && strcmp(str, "halt") == 0)
1210 cpu_reboot(RB_HALT, NULL);
1211 else if (len == 6 && strcmp(str, "reboot") == 0)
1212 cpu_reboot(0, NULL);
1213 #if defined(DDB)
1214 else if (len == 3 && strcmp(str, "ddb") == 0)
1215 console_debugger();
1216 #endif
1217
1218 cp = str + len - 1;
1219 c = *cp;
1220
1221 if ((wname = getwedgename(str, len)) != NULL) {
1222 if ((dv = dkwedge_find_by_wname(wname)) == NULL)
1223 return NULL;
1224 part = defpart;
1225 goto gotdisk;
1226 } else if (c >= 'a' && c <= ('a' + MAXPARTITIONS - 1)) {
1227 part = c - 'a';
1228 *cp = '\0';
1229 } else
1230 part = defpart;
1231
1232 #ifdef MEMORY_DISK_HOOKS
1233 for (i = 0; i < NMD; i++)
1234 if (strcmp(str, fakemdrootdev[i].dv_xname) == 0) {
1235 dv = &fakemdrootdev[i];
1236 goto gotdisk;
1237 }
1238 #endif
1239
1240 dv = finddevice(str);
1241 if (dv != NULL) {
1242 if (device_class(dv) == DV_DISK) {
1243 gotdisk:
1244 majdev = devsw_name2blk(dv->dv_xname, NULL, 0);
1245 if (majdev < 0)
1246 panic("parsedisk");
1247 if (DEV_USES_PARTITIONS(dv))
1248 *devp = MAKEDISKDEV(majdev, device_unit(dv),
1249 part);
1250 else
1251 *devp = makedev(majdev, device_unit(dv));
1252 }
1253
1254 if (device_class(dv) == DV_IFNET)
1255 *devp = NODEV;
1256 }
1257
1258 *cp = c;
1259 return (dv);
1260 }
1261
1262 /*
1263 * snprintf() `bytes' into `buf', reformatting it so that the number,
1264 * plus a possible `x' + suffix extension) fits into len bytes (including
1265 * the terminating NUL).
1266 * Returns the number of bytes stored in buf, or -1 if there was a problem.
1267 * E.g, given a len of 9 and a suffix of `B':
1268 * bytes result
1269 * ----- ------
1270 * 99999 `99999 B'
1271 * 100000 `97 kB'
1272 * 66715648 `65152 kB'
1273 * 252215296 `240 MB'
1274 */
1275 int
1276 humanize_number(char *buf, size_t len, uint64_t bytes, const char *suffix,
1277 int divisor)
1278 {
1279 /* prefixes are: (none), kilo, Mega, Giga, Tera, Peta, Exa */
1280 const char *prefixes;
1281 int r;
1282 uint64_t umax;
1283 size_t i, suffixlen;
1284
1285 if (buf == NULL || suffix == NULL)
1286 return (-1);
1287 if (len > 0)
1288 buf[0] = '\0';
1289 suffixlen = strlen(suffix);
1290 /* check if enough room for `x y' + suffix + `\0' */
1291 if (len < 4 + suffixlen)
1292 return (-1);
1293
1294 if (divisor == 1024) {
1295 /*
1296 * binary multiplies
1297 * XXX IEC 60027-2 recommends Ki, Mi, Gi...
1298 */
1299 prefixes = " KMGTPE";
1300 } else
1301 prefixes = " kMGTPE"; /* SI for decimal multiplies */
1302
1303 umax = 1;
1304 for (i = 0; i < len - suffixlen - 3; i++)
1305 umax *= 10;
1306 for (i = 0; bytes >= umax && prefixes[i + 1]; i++)
1307 bytes /= divisor;
1308
1309 r = snprintf(buf, len, "%qu%s%c%s", (unsigned long long)bytes,
1310 i == 0 ? "" : " ", prefixes[i], suffix);
1311
1312 return (r);
1313 }
1314
1315 int
1316 format_bytes(char *buf, size_t len, uint64_t bytes)
1317 {
1318 int rv;
1319 size_t nlen;
1320
1321 rv = humanize_number(buf, len, bytes, "B", 1024);
1322 if (rv != -1) {
1323 /* nuke the trailing ` B' if it exists */
1324 nlen = strlen(buf) - 2;
1325 if (strcmp(&buf[nlen], " B") == 0)
1326 buf[nlen] = '\0';
1327 }
1328 return (rv);
1329 }
1330
1331 /*
1332 * Return true if system call tracing is enabled for the specified process.
1333 */
1334 bool
1335 trace_is_enabled(struct proc *p)
1336 {
1337 #ifdef SYSCALL_DEBUG
1338 return (true);
1339 #endif
1340 #ifdef KTRACE
1341 if (ISSET(p->p_traceflag, (KTRFAC_SYSCALL | KTRFAC_SYSRET)))
1342 return (true);
1343 #endif
1344 #ifdef SYSTRACE
1345 if (ISSET(p->p_flag, PK_SYSTRACE))
1346 return (true);
1347 #endif
1348 #ifdef PTRACE
1349 if (ISSET(p->p_slflag, PSL_SYSCALL))
1350 return (true);
1351 #endif
1352
1353 return (false);
1354 }
1355
1356 /*
1357 * Start trace of particular system call. If process is being traced,
1358 * this routine is called by MD syscall dispatch code just before
1359 * a system call is actually executed.
1360 * MD caller guarantees the passed 'code' is within the supported
1361 * system call number range for emulation the process runs under.
1362 */
1363 int
1364 trace_enter(struct lwp *l, register_t code,
1365 register_t realcode, const struct sysent *callp, void *args)
1366 {
1367 #if defined(SYSCALL_DEBUG) || defined(KTRACE) || defined(PTRACE) || defined(SYSTRACE)
1368 struct proc *p = l->l_proc;
1369
1370 #ifdef SYSCALL_DEBUG
1371 scdebug_call(l, code, args);
1372 #endif /* SYSCALL_DEBUG */
1373
1374 ktrsyscall(code, realcode, callp, args);
1375
1376 #ifdef PTRACE
1377 if ((p->p_slflag & (PSL_SYSCALL|PSL_TRACED)) ==
1378 (PSL_SYSCALL|PSL_TRACED))
1379 process_stoptrace(l);
1380 #endif
1381
1382 #ifdef SYSTRACE
1383 if (ISSET(p->p_flag, PK_SYSTRACE)) {
1384 int error;
1385 KERNEL_LOCK(1, l);
1386 error = systrace_enter(l, code, args);
1387 KERNEL_UNLOCK_ONE(l);
1388 return error;
1389 }
1390 #endif
1391 #endif /* SYSCALL_DEBUG || {K,P,SYS}TRACE */
1392 return 0;
1393 }
1394
1395 /*
1396 * End trace of particular system call. If process is being traced,
1397 * this routine is called by MD syscall dispatch code just after
1398 * a system call finishes.
1399 * MD caller guarantees the passed 'code' is within the supported
1400 * system call number range for emulation the process runs under.
1401 */
1402 void
1403 trace_exit(struct lwp *l, register_t code, void *args, register_t rval[],
1404 int error)
1405 {
1406 #if defined(SYSCALL_DEBUG) || defined(KTRACE) || defined(PTRACE) || defined(SYSTRACE)
1407 struct proc *p = l->l_proc;
1408
1409 #ifdef SYSCALL_DEBUG
1410 scdebug_ret(l, code, error, rval);
1411 #endif /* SYSCALL_DEBUG */
1412
1413 ktrsysret(code, error, rval);
1414
1415 #ifdef PTRACE
1416 if ((p->p_slflag & (PSL_SYSCALL|PSL_TRACED)) ==
1417 (PSL_SYSCALL|PSL_TRACED))
1418 process_stoptrace(l);
1419 #endif
1420
1421 #ifdef SYSTRACE
1422 if (ISSET(p->p_flag, PK_SYSTRACE)) {
1423 KERNEL_LOCK(1, l);
1424 systrace_exit(l, code, args, rval, error);
1425 KERNEL_UNLOCK_ONE(l);
1426 }
1427 #endif
1428 #endif /* SYSCALL_DEBUG || {K,P,SYS}TRACE */
1429 }
1430
1431 /*
1432 * Disable kernel preemption.
1433 */
1434 void
1435 crit_enter(void)
1436 {
1437 /* nothing */
1438 }
1439
1440 /*
1441 * Reenable kernel preemption.
1442 */
1443 void
1444 crit_exit(void)
1445 {
1446 /* nothing */
1447 }
1448